Q&A: ABB's Chief Technology Officer on Why Robotics Is Entering Its 'iPhone Moment'

We sat down with Marc Segura, President of ABB Robotics, to discuss how foundation models are changing what robots can do — and what it means for manufacturers who've been waiting for automation to get easier.

Industry 4.1: ABB has been talking about AI-enabled robotics for years. What's actually different now?

Marc Segura: The honest answer is that until about 18 months ago, most of what we called "AI-enabled robotics" was really just better software — improved path planning, faster teach-pendant programming, better sensor integration. Useful, but incremental.

What's different now is foundation models. We're seeing robot systems that can generalize from limited training data in ways that were simply impossible before. Our new OmniCore controller, combined with large vision-language models, can handle tasks that previously required thousands of hours of programming. A robot can now look at a new part it's never seen before, understand its geometry, figure out how to grasp it, and place it correctly — with maybe 10 minutes of demonstration instead of 10 days of programming.

I4.1: You've called this an "iPhone moment" for robotics. That's a big claim.

Segura: I use that analogy because the iPhone didn't invent the touchscreen or the mobile browser. What it did was package existing technologies into something that was genuinely easy to use. That's what's happening in robotics right now.

We have a customer in Sweden — a contract manufacturer doing electronics assembly. They need to change their production setup 3–4 times per week for different customers. Under the old model, each changeover required a programmer to spend 6–8 hours reprogramming robot paths, adjusting vision parameters, updating pick-and-place coordinates. That made automation economically questionable for their business.

With our new system, the line operator — not a programmer — shows the robot the new task. Literally demonstrates it with their hands while the robot watches. The system generalizes from that demonstration and starts running within 30 minutes. Their changeover time went from 8 hours to 30 minutes. That's the iPhone moment — when the technology disappears and the capability just works.

I4.1: What's the realistic timeline for this to be mainstream? Factory floor conditions aren't exactly controlled lab environments.

Segura: You're right, and that's an important caveat. The demonstrations work beautifully in structured environments — bin picking, assembly, palletizing. For unstructured environments — think a field service robot working in a refinery, or a construction robot on a job site — we're still 3–5 years away from reliable deployment.

For manufacturing specifically, I'd say the technology is production-ready today for 60–70% of typical assembly and material handling tasks. The remaining 30–40% — tasks requiring very high precision, force-sensitive operations, or handling highly deformable materials — still need traditional programming approaches, though that's shrinking fast.

I4.1: The elephant in the room: what happens to the jobs?

Segura: I know this is the question everyone wants to ask. Here's what our data actually shows: our top 50 customers by robot installations have collectively increased their manufacturing workforce by 12% over the past five years while nearly doubling their robot density. The robots aren't replacing people — they're enabling growth that wouldn't be possible otherwise.

The caveat is that the jobs are changing. Fewer people doing repetitive manual tasks, more people doing programming, supervision, maintenance, and quality oversight. The companies that handle that transition well — with real training programs, not just lip service — are the ones whose workers embrace automation rather than fear it.

I4.1: What should a plant manager who's never deployed a robot do first?

Segura: Start with palletizing. I know that sounds boring, but it's boring for a reason — it's the most proven, lowest-risk robotic application in manufacturing. Nearly every plant has palletizing tasks. The ROI is clear and measurable. The technology is mature. And it gives your team experience working with robots before you tackle more complex applications.

The biggest mistake I see is companies starting with their most complex process because that's where the biggest ROI potential is. They're right about the potential, but the risk is too high for a first deployment. Start simple, build capability, then go after the complex stuff.

This interview has been edited for length and clarity. Elena Vasquez covers production and robotics for Industry 4.1.